The whole point is there being pixels so small you can’t see the discrete steps in curves of text etc. that’s a different problem to not being able to see discrete pixels.
The resolving power of the eye is usually given as 1/60 of a degree for a person with 6/6 (20/20 in bushels per fortnight land) vision. That's the ability to just see two black lines, set 1/60 of a degree apart, with white in between them, rather than perceive a single grey line. To model this as pixels, the pixels would need to subtend 1/120 of a degree (to get black/white/black with the black 1/60º apart), which at 1m would make a pixel 145 um wide, or 72.7 um at 0.5 m (or 349 pixels/inch at 0.5m).
However, we can distinguish discontinuities at a finer level. So draw a line, cut through it and offset one half from the other and we can spot that at even smaller scale. (Haven't got hard figures to hand)
Vision is complicated, and pixels/inch numbers is too simplistic to describe what can and can't be seen. Other things come into the equation, contrast, illumination level and so on. But if you have to decide on a pixel size that's close to the limits of human vision then 1/120 of a degree subtended angle would be the number to pick.